Free
Research Article  |   October 2008
Fixational eye movements predict the perceived direction of ambiguous apparent motion
Author Affiliations
Journal of Vision October 2008, Vol.8, 13. doi:10.1167/8.14.13
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to Subscribers Only
      Sign In or Create an Account ×
    • Get Citation

      Jochen Laubrock, Ralf Engbert, Reinhold Kliegl; Fixational eye movements predict the perceived direction of ambiguous apparent motion. Journal of Vision 2008;8(14):13. doi: 10.1167/8.14.13.

      Download citation file:


      © 2017 Association for Research in Vision and Ophthalmology.

      ×
  • Supplements
Abstract

Neuronal activity in area LIP is correlated with the perceived direction of ambiguous apparent motion (Z. M. Williams, J. C. Elfar, E. N. Eskandar, L. J. Toth, & J. A. Assad, 2003). Here we show that a similar correlation exists for small eye movements made during fixation. A moving dot grid with superimposed fixation point was presented through an aperture. In a motion discrimination task, unambiguous motion was compared with ambiguous motion obtained by shifting the grid by half of the dot distance. In three experiments we show that (a) microsaccadic inhibition, i.e., a drop in microsaccade frequency precedes reports of perceptual flips, (b) microsaccadic inhibition does not accompany simple response changes, and (c) the direction of microsaccades occurring before motion onset biases the subsequent perception of ambiguous motion. We conclude that microsaccades provide a signal on which perceptual judgments rely in the absence of objective disambiguating stimulus information.

Introduction
Multistable perceptual phenomena such as alternations in visual awareness without accompanying changes of the visual stimulus have been of interest to psychologists and physiologists for centuries (e.g., Leopold & Logothetis, 1999; von Helmholtz, 1867), and some of the associated stimuli have percolated into popular culture. Yet the mechanisms underlying these seemingly spontaneous alternations have not been fully identified. Traditional explanations for perceptual reversals involve competitive interactions within the visual system. However, since the beginning of research on perceptual alternations, eye movements have also been thought to play a role. Although bottom-up signals following small eye movements have been suggested to be responsible for perceptual alternations, this is an unresolved research problem (Kanai, Moradi, Shimojo, & Verstraten, 2005). Here we present (a) clear evidence that small eye movements made during fixation are correlated with perceptual changes and (b) tentative evidence for a causal influence of microsaccades on the perceived state of ambiguous apparent motion. 
A defining characteristic of bistable visual stimuli is that an observer experiences transitions between two percepts in the absence of corresponding changes in the stimulus. For example, ambiguous apparent motion stimuli (e.g., Ramachandran & Anstis, 1983) lead to the relatively stable percept that the stimulus is moving into a single direction, whereas physical motion direction is ambiguous. If physical stimulus properties cannot be responsible for the resolution of ambiguity, what then determines perceived direction? Obviously, it must be caused by the state of some internal (cognitive or physiological) variable. For example, stochastic activity in some perception-related brain area before or at the time of motion presentation could bias perception in one way or the other. 
However, “stochastic activity” can often be a convenient umbrella term for unknown influences. Eye movements may be one such influence. To this end, van Dam and van Ee (2005, 2006a) have shown that retinal image shifts caused by saccades can cause alternations in awareness during binocular rivalry. Specifically, these authors made a division between trials in which saccades did vs. did not lead to a change in the retinal image of an oblique sinusoidal grating (i.e., half-cycle vs. full-cycle shifts, respectively). Using an analysis time-locked to perceptual flips, they could show that only the frequency of saccades that caused a change in the retinal image is significantly increased before flips. In addition, and independent of retinal change, perceptual changes were always followed by a drop in saccade frequency occurring around the time of the report of the flip. 
We suspect that fixational eye movements, which have been shown to affect perception in other domains (e.g., Martinez-Conde, Macknik, & Hubel, 2002; Martinez-Conde, Macknik, Troncoso, & Dyar, 2006), also contribute to alternations in bistable motion perception. During a fixation, the eye is not completely still. Several types of small eye movements occur; generally they are less than one degree in size. Stretches of slow drift are followed by sudden, tiny, saccade-like jumps called microsaccades (see Engbert, 2006; Martinez-Conde, Macknik, & Hubel, 2004 for recent reviews). Here we investigate whether fixational eye movements are correlated with changes in the perceived direction of ambiguous apparent motion. Given the appropriate temporal relationship, such a correlation would not only encourage the use of eye movements as a proxy measure of perceptual state, but also suggest that eye movements may cause changes in the perceptual state. 
One ambiguous apparent motion stimulus that elicits a vivid sensation of motion into one direction consists of a simple alternation of two grids with suitable timing, related to each other by a shift of half of the grid's period. Although the perceived motion direction is random from trial to trial, the perception is relatively stable in time within a trial, before it flips (typically after more than a second). Williams, Elfar, Eskandar, Toth, and Assad (2003) used such a stimulus and recorded single-cell activity from areas known to be involved in motion perception. They found that neurons in the lateral intraparietal area (LIP) responded more strongly on trials in which the perception was compatible with the neurons' preferred direction. This selectivity was less common in the medial superior temporal area (MST) and virtually absent in the middle temporal area (MT). Thus, the more upstream an area specialized in motion perception was, the less it was affected by decisional biases. Importantly, variations in activity of MST and especially LIP neurons just before motion onset were predictive of the subsequently perceived direction. 
Because LIP is not a primary perceptual area but is also involved in eye movement planning (Barash, Bracewell, Fogassi, Gnadt, & Andersen, 1991), sensorimotor transformations (Zhang & Barash, 2000, 2004), and spatial attention (Bisley & Goldberg, 2003; Gottlieb, 2007), we thought it likely that some correlate of a perceptual decision should also be detectable in fixational eye movements, using a paradigm modeled on Williams et al. (2003). These authors were primarily interested in the neural correlate of perceptual flips; they did not report detailed analyses of fixational eye movements but just checked for differences in microsaccade direction over the whole trial. Therefore, they were unlikely to pick up temporary modulations. Eye movements can be used as an online indicator of the state of perception: for example, Madelain and Krauzlis (2003) reported that smooth pursuit eye movement followed the perception of motion direction with an average delay of 53 ms. They can also influence perception (Collins & Doré-Mazars, 2006). With regard to motion stimuli, it has been shown that saccades can (a) enhance perception of certain classes of (static and) moving stimuli, namely static gratings with low spatial frequency and dynamic gratings oscillating with high temporal frequencies, and (b) have an enhanced likelihood of occurrence before a perceptual decision (Deubel & Elsner, 1986). 
There is amassing evidence that eye movements and attention are closely linked (Deubel & Schneider, 1996; Hoffman & Subramaniam, 1995). Such links have also been reported for fixational microsaccades (Engbert & Kliegl, 2003; Hafed & Clark, 2002; Laubrock, Engbert, & Kliegl, 2005; Laubrock, Engbert, Rolfs, & Kliegl, 2007). Because attention is often operationally defined by enhanced perceptual performance, microsaccades might also be expected to correlate with perception. Indeed, extending earlier reports (Steinman & Cunitz, 1968), it has recently been shown that fixational eye movements are correlated with disappearance and reappearance of the percept in perceptual fading, e.g., Troxler fading (Martinez-Conde et al., 2006), or with perceptual state changes in binocular rivalry (van Dam & van Ee, 2005, 2006a). 
Does such a correlation also exist between microsaccades and perceptual flips of ambiguous apparent motion? To determine whether fixational eye movements are related to motion perception, we measured eye movements in a task closely modeled after the one used by Williams et al. (2003). First, in Experiment 1 and a control experiment, we investigated whether fixational eye movements are temporally correlated to perceptual flips. In Experiment 2, we looked at whether fixational eye movements might have a causal role for perceptual flips, and also differentiated between the influence of spontaneous and attention-induced microsaccades. 
Experiment 1
In the first experiment, we presented an apparent motion stimulus for a relatively long period and asked participants to continuously report the perceived direction of motion. We contrasted unambiguous motion (with pre-programmed changes in motion direction) with ambiguous motion (where changes in perceived motion direction occur spontaneously). A control experiment was designed to rule out that an eventual correlation was purely due to stimulus or response demands of the task. 
Method
Participants
Sixteen members of the University of Potsdam subject pool (14 female, 2 male, mean age: 25.2, range 19–41 years) participated in the experiment, which lasted for about 45 minutes. Participants had the option to receive either course credit or seven euros for their participation. Another 16 participants (12 f, 4 m, mean age: 24.1, range 19–30 years) from the same pool were recruited for a control experiment, reported below. 
Apparatus and materials
Eye movements were recorded binocularly with the video-based EyeLink II system (SR research, Toronto, Canada) at a sampling rate of 500 Hz. Stimuli were displayed on a 19″ Eye-Q 650 CRT at a resolution of 1024 × 768 and a refresh rate of 100 Hz. Head movements were restrained and a viewing distance of 50 cm was assured by use of a chin rest. A Power Macintosh G4 running Mac OS 9.2 controlled stimulus presentation and eye movement recording by communicating via an Ethernet TCP/IP direct link with an Intel Pentium 4 based Dell Optiplex system containing the eye tracker card. The experimental software controlling stimulus display and response collection was implemented in Matlab, using the Psychophysics (Brainard, 1997; Pelli, 1997) and EyeLink (Cornelissen, Peters, & Palmer, 2002) toolboxes. All timing of display events was based on the vertical retrace interrupt. Responses were collected using the left and right arrow keys on a standard USB keyboard. 
All screens featured a dark gray background (2.5 cd/m 2) and a central circular aperture (15 deg diameter) with a black background (0.25 cd/m 2). The motion pattern was a grating consisting of the points (point size: one pixel, point color: white, 88 cd/m 2) located at the crossings of a rectangular mesh grid with a mesh width of 0.8 degrees (20 pixels). In the center and on top of the grid, a superimposed fixation spot composed of a light gray (16.7 cd/m 2) disk (0.6 deg diameter) with a central black dot (0.15 deg diameter) was permanently displayed. The grid was larger than the aperture diameter, but only grid points that fell within the aperture were visible. 1 shows a downscaled version of the animated stimulus, and Figure 1 illustrates the most important grid parameters in a space-time diagram. 
 
Movie 1
 
Ambiguous apparent motion stimulus (not drawn to scale). Double click to show animation.
Figure 1
 
Characteristics of the moving grid.
Figure 1
 
Characteristics of the moving grid.
Procedure
At the beginning of each trial, a drift correction was performed using a drift correction target with visual properties identical to the fixation spot. If the drift correction was unsuccessful, the eye tracker was re-calibrated. Normal re-calibrations were performed after every 8th trial. Following successful drift correction, an initial fixation screen showing a static grating inside the aperture was displayed. 
After a fixed interval of 1500 ms during which participants were required to fixate, the grid started to move horizontally. In the unambiguous conditions, motion direction changed (from left to right or vice versa) after random intervals independently drawn from a 9th order gamma distribution with an expected value of 1000 ms, plus an added constant of 300 ms. The number of changes in motion direction was pre-programmed to amount to a minimum total trial duration of 20 seconds (range 20.2–23.9 s). For trials with ambiguous motion, the same number of intervals were randomly drawn (and actually, the same changes in sign of motion direction were programmed, but without visible effect), thus total trial duration was equivalent to unambiguous trials. 
Participants were told that they would see a grating that would start to move and that could change direction several times during a trial. They were instructed to start responding after motion onset and to continuously depress the key corresponding to the momentarily perceived motion direction (i.e., left and right arrow key for motion to the left and right, respectively), while keeping their gaze at the fixation point. 
Design
A session consisted of 72 experimental and 4 warmup trials and lasted for about 45 minutes. Four combinations of motion speed and pixel displacements between frames were used, leading to two conditions with unambiguous and two conditions with ambiguous motion. Three quarters of the trials used a velocity of 3.0 deg/s. An equal amount of these trials used displacements between frames of 8, 9, and 10 pixels between visible dots, which were spaced 20 pixels apart. The ambiguous condition (10 pixels dot spacing) was additionally tested at a higher velocity of 4.5 deg/s in the remaining quarter of trials to check whether a change in velocity in the range tested would have any impact on micro-movements. 
For ease of report, we will use the following abbreviations (corresponding to the displacement in pixels of the grid and an added tag to describe ambiguous motion speed) for the stimulus conditions: 8/20 for unambiguous-easy, 9/20 unambiguous-difficult, 10/20-normal for ambiguous motion at 3 deg/s, and 10/20-fast for ambiguous motion at 4.5 deg/s. 
The different levels of ambiguity differed in the degree of displacement of the grid between frames. For example, in the ambiguous 10/20 condition, the grid was displaced by 0.4 degrees between frames, resulting in an alternation of two grids at a rate of 6.67 Hz for 10/20-normal, whereas in the unambiguous 8/20 condition, the grid was displaced by less than half the dot spacing (0.35 deg). 
Because velocity was kept constant, the rate of display changes was somewhat higher in the unambiguous 9/20 and 8/20 conditions (7.52 and 8.54 Hz, respectively). The higher velocity in the ambiguous 10/20-fast condition implied a frame rate of 10 Hz. 
Control experiment
A control experiment was conducted to rule out that eventual effects were purely related to response preparation and execution. The visual display was exactly the same as in the original experiment (but only the ambiguous condition at 3.0 deg/s) on half of the trials; on the other half of the trials, after 1500 ms an initial static grid inside the aperture was cleared so that an empty aperture (except for the fixation dot) stayed on screen for the remainder of the trial duration. Demands on response execution were also identical to the original experiment, as the same set of keyboard responses was used. However, the task was changed. Instead of having to report the state of motion perception, participants in the control experiment were told that the task was to judge subjective temporal intervals and instructed to alternately press and hold the left and right key after they felt a second had passed. Trial duration was as in the original experiment. 
Data analysis
Invalid trials due to recording errors or an excessive amount of eye blinks (more than 100 ms) were discarded. Overall, this led to the rejection of 226 trials (median N of missing trials per subject: 7.5). Due to the long trial duration, it turned out to be impractical to completely remove trials with blinks. Instead, stretches with 40 ms padding before and after the occasional within-trial blink were deleted from the time series and discarded from further analysis. 
Microsaccades were detected by application of the algorithm described by Engbert and Kliegl (2003), with parameters chosen so that microsaccades are outliers in velocity space (characterized by a binocular deviation from mean velocity of more than six-fold of a robust standard deviation estimate during at least four consecutive samples). The upper limit of microsaccade size was set to one degree; the modal amplitude of microsaccades detected was 24.9 arcmin and hence higher than the roughly 14 arcmin obtained in this laboratory during fixation tasks with a static background (e.g., Engbert, 2006). 
Each microsaccade was characterized by its moment of occurrence (halfway between onset and offset), direction (from onset to offset), and congruency of its direction with the perceived direction of motion. Independent of its moment of occurrence, a microsaccade was classified as perception-congruent if the horizontal component of its direction was the same as the momentarily perceived motion direction and as incongruent otherwise. Microsaccade rates and congruency effects were calculated by selecting trial events, selecting epochs surrounding these events, averaging over events in a given condition for each subject, and then averaging over subjects. Depending on the focus of the analysis, these trial events could either be changes in reported perceived direction (in both ambiguous and unambiguous conditions) for response-aligned analyses or pre-programmed changes in motion direction (in the unambiguous condition) or else the temporally closest corresponding frame onset leading to a display change (in the ambiguous conditions, where there were no physical changes of motion direction) for stimulus-locked analyses. 
The significance level of statistical tests was set to 0.05. In case of multiple tests, to protect from spurious results, significance levels were adjusted according to the false-discovery rate procedure (Benjamini & Hochberg, 1995). 
Results
Summary statistics
The mean interval between perceptual flips and hence the number of perceptual flips depended on the stimulus condition (see Table 1). The frequency of perceptual flips in the ambiguous motion conditions was smaller than expected, indicating that perceived motion direction is stable for more than 2 seconds. The mean interval between two microsaccades was similar in all stimulus conditions, with a grand mean of 1098 ms and considerable variability both within and between subjects (not shown). The distribution of the mean interval between microsaccades is unimodal and positively skewed and therefore similar to the shape of distribution of periods between perceptual alternations in rivalry. We compared maximum likelihood fits of gamma, inverse gamma, and lognormal distributions to the distribution of the mean inter-microsaccade interval, and the lognormal (mean = 6.43, SD = 0.7 on a log scale) turned out to be clearly superior to the other two ( Figure 2). 
Table 1
 
Statistics (between-subject means and standard errors in milliseconds) for interval between subsequent perceptual flips, interval between subsequent microsaccades, and “reaction time” to changes in stimulus direction in Experiment 1.
Table 1
 
Statistics (between-subject means and standard errors in milliseconds) for interval between subsequent perceptual flips, interval between subsequent microsaccades, and “reaction time” to changes in stimulus direction in Experiment 1.
Condition Perceptual flip waiting time Microsaccades Reaction time
Mean SE Mean SE Mean SE
Unambiguous, 8/20 1468 28.2 1046 105.2 595 33.4
Unambiguous, 9/20 1504 38.6 1126 100.9 672 33.2
Ambiguous 10/20 3077 328.2 1131 102.6 N/A
Ambiguous 10/20, fast 4591 613.1 1113 110.7 N/A
Figure 2
 
Distribution of intervals between consecutive microsaccades in Experiment 1. Left: Fit of lognormal, gamma, and inverse gamma distributions to the empirical density (black), smoothed with a Gaussian kernel ( SD 64.8 ms). Right: Empirical vs. fitted 150-tiles in a log–log coordinate system. Both plots show that although there are some systematic deviations, the lognormal evidently fits better than the gamma or inverse gamma distributions.
Figure 2
 
Distribution of intervals between consecutive microsaccades in Experiment 1. Left: Fit of lognormal, gamma, and inverse gamma distributions to the empirical density (black), smoothed with a Gaussian kernel ( SD 64.8 ms). Right: Empirical vs. fitted 150-tiles in a log–log coordinate system. Both plots show that although there are some systematic deviations, the lognormal evidently fits better than the gamma or inverse gamma distributions.
Within the relatively long trials, directions of consecutive microsaccades were not independent, but negatively correlated, e.g., a microsaccade to the right was likely to be followed by a microsaccade to the left. The deviation from independence was particularly pronounced for ambiguous motion: the log odds ratios were −0.43, −0.62, −0.86, and −0.85 for the 8/20, 9/20, 10/20, and 10/20-fast conditions, respectively. Confidence intervals based on asymptotic standard errors excluded zero for all four conditions, and confidence intervals for either of the ambiguous conditions did not overlap with either of the unambiguous conditions. 
The reaction time between a pre-programmed change in motion direction and the corresponding change in response also depended on the stimulus condition: it was longer in the 9/20 than in the 8/20 condition, indicating that motion changes in the closer-to-ambiguous motion condition were indeed more difficult to perceive. 
Microsaccade rate aligned to changes in reported perceived movement direction
When the moment of occurrence of microsaccades is plotted relative to the time of report of a perceptual flip, a clear correlation is apparent in the global modulation of microsaccade rate on both ambiguous and unambiguous trials ( Figure 3A). Microsaccade frequency drops well before a reported flip and then increases above the baseline level around the time of the report. Taking reaction time into account, it appears that the earliest signs of a drop in microsaccade frequency are detectable around the time of a perceptual flip, about 500 ms before the actual response. 
Figure 3
 
Microsaccade rates in Experiment 1; (A) response-locked and (B) stimulus-locked. Microsaccadic inhibition is locked to task-relevant perceptual changes (A). Additionally, every frame-onset causes microsaccadic inhibition with a certain probability (the oscillatory modulation in (B)).
Figure 3
 
Microsaccade rates in Experiment 1; (A) response-locked and (B) stimulus-locked. Microsaccadic inhibition is locked to task-relevant perceptual changes (A). Additionally, every frame-onset causes microsaccadic inhibition with a certain probability (the oscillatory modulation in (B)).
Note that this result is also obtained for ambiguous motion (here the average reaction time to real stimulus switches can be used as a somewhat liberal estimate of when the flip actually occurred, see van Dam & van Ee, 2005), although the curves do not look as smooth as for unambiguous motion simply because of the smaller number of responses made (compare Table 1). Thus, changes in microsaccade frequency appear to reflect perceptual rather than purely physical changes. 
Microsaccade rate aligned to stimulus changes
Plotting the moment of occurrence of microsaccades aligned to changes in motion direction (for unambiguous motion) or to the closest corresponding frame change (for ambiguous motion) reveals an oscillatory part and a more global modulation visible as a trend in the offset around which the data oscillate ( Figure 3B). Thus, a global rate modulation related to perceptual flips is observed like in the response-aligned analysis. The rate inhibition preceding perceptual flips appears to coincide with the time of change in motion direction. On ambiguous motion trials, there is no global modulation simply because there is no physical change in motion direction. Note the contrast to the response-locked analysis. Taken together, this suggests that the global modulation is caused by any changes in motion direction, whether perceived or real. 
The second result is a local, oscillatory modulation of microsaccade frequency that is independent of stimulus ambiguity. The frequency of this local modulation is closely coupled to the temporal inter-change interval between frames. Remember that to achieve constant stimulus velocity, different inter-frame intervals were chosen for 10/20 vs. 9/20 vs. 8/20 pixel replacements. These different intervals are reflected in the period of the local modulation. For a formal evaluation, individual stimulus-onset aligned, time-dependent microsaccade occurrence histograms were smoothed with a rectangular window of width 50 ms. The average across subjects was analyzed by applying a discrete Fourier transform. The maxima (and clear outliers) of the detrended, log-transformed power spectrum are listed in the “microsaccades” column of Table 2, which shows that the estimated peaks very closely followed the frame rate. The local modulation thus indicates that each frame change in a motion stimulus (in the range of frame rates tested here) induces a tendency for microsaccade inhibition. Almost identical results were obtained when the first two stimulus changes per trial were excluded from the analysis. Therefore, results are not due to a propagation of the initial, motion-onset-related inhibition. 
Table 2
 
Frame rate of stimulus changes in the different conditions (Hz) and rate of local modulations of microsaccade rate (Hz).
Table 2
 
Frame rate of stimulus changes in the different conditions (Hz) and rate of local modulations of microsaccade rate (Hz).
Condition Stimulus Microsaccades
Unambiguous, 8/20 8.54 8.54
Unambiguous, 9/20 7.52 7.48
Ambiguous 10/20, normal 6.67 6.62
Ambiguous 10/20, fast 10.00 9.95
Micro-OKN?
Whole-field moving grating stimuli are known to induce optokinetic nystagmus (OKN). OKN—a periodic, sawtooth-like pattern consisting of an alternation of pursuit-like slow phases and saccade-like fast phases in the opposite direction—is typically regarded as a phylogenetically old reflex that is observed in all animals with moving eyes. Our stimulus was not whole field, and OKN is greatly reduced under visual fixation conditions. Therefore, it comes to no surprise that no obvious sawtooth-like OKN pattern was present in our data. However, if the data are analyzed at a level of high resolution, a signature of statistical OKN might be found. To do this, we defined an ad hoc measure of “systematic drift”: for stretches between microsaccades, the time series was low-pass filtered and then the median distance traveled was tabulated and assigned to the center point of time of the interval. With this table, an event-related analysis similar to the microsaccade analysis was performed. 
Event-related averaging of systematic drift shows that the drift phases between microsaccades are biased: drift direction tends to coincide with the direction of stimulus motion. Drift direction is also negatively correlated with the direction of the subsequent microsaccades. Thus, the stimulus-correlated pattern we observed here resembles OKN, but it occurs on a “microscopic” scale and with a less regular (i.e., only statistical) pattern. It is like a statistical micro-OKN, consisting of microsaccades and stretches of systematic drift. Neither of these occurs very frequently or regularly, but if they do occur they tend to be stimulus related. 
Control experiment
Microsaccade rate was somewhat higher, and the response buttons were pressed somewhat more often (reflecting the task to alternate key presses every second) in the control experiment, where the task had been changed from report of motion direction to temporal interval estimation ( Table 3, compare with Table 1). Temporal interval estimates were close to the true duration of one second, but it was systematically underestimated, and more so with a moving grid than with an empty aperture. The correlation of subsequent microsaccade directions was again negative. Interestingly, it did not differ in size between whether a moving grid was visible or not. 
Table 3
 
Statistics (between-subject means and standard errors in milliseconds) for intervals between subsequent responses and intervals between subsequent microsaccades in the control experiment.
Table 3
 
Statistics (between-subject means and standard errors in milliseconds) for intervals between subsequent responses and intervals between subsequent microsaccades in the control experiment.
Condition Response alternations Microsaccades
Mean SE Mean SE
Moving grid (ambiguous, 10/20) 945 6.5 926 115.9
No stimulus 985 5.9 849 105.3
Whereas the stimulus-related modulations of microsaccade rate are clearly obtained again (in the motion condition, Figure 4B), there are no response-related modulations of microsaccade rate in either the motion or the “no stimulus” condition ( Figure 4A). Comparison of rate modulations in the two experiments suggests that microsaccadic inhibition preceding responses in the original experiment was not due to responding per se but likely related to perceptual decisions. 
Figure 4
 
Microsaccade rates in the control experiment comparing temporal judgment responses during presentation of a moving vs. static grid, (A) response-locked and (B) stimulus-locked. If changes are not task relevant, no global response-locked microsaccadic inhibition is observed (compare (A) with Figure 3A). However, the moving grid still (as in Figure 3B) elicits frame-locked microsaccadic inhibition, visible in the oscillatory pattern in (B).
Figure 4
 
Microsaccade rates in the control experiment comparing temporal judgment responses during presentation of a moving vs. static grid, (A) response-locked and (B) stimulus-locked. If changes are not task relevant, no global response-locked microsaccadic inhibition is observed (compare (A) with Figure 3A). However, the moving grid still (as in Figure 3B) elicits frame-locked microsaccadic inhibition, visible in the oscillatory pattern in (B).
Discussion
Microsaccade rate aligned to stimulus changes
Conditions for ambiguous apparent motion required a relatively slow frame rate of ca. 7 Hz, clearly below the flicker fusion frequency. However, due to the dark background and the relatively small dot grids, motion did not phenomenologically appear to be very jumpy. Therefore, despite the documented sensitivity of the visual system to flicker (for example, neuronal activity in LGN even follows the flicker associated with a 74-Hz refresh rate of the monitor; Martinez-Conde et al., 2002), we initially found it quite astonishing that frame onsets were so noticeable that each had a chance to be reflected in fixational eye movements. 
However, optokinetic nystagmus (OKN) is a well-known phenomenon that is also caused by moving grating-like stimuli. Although OKN is greatly reduced under visual fixation, the stimulus-correlated pattern we observed here resembles OKN on a more microscopic scale. Direction of the slow phase of OKN has been found to be a reliable indicator of the perceived direction of motion in rivalry stimuli, when oppositely moving drifting gratings are presented to each eye (Enoksson, 1963, 1968). Thus, it has been suggested that OKN can be used as an “objective indicator of the subject's perception” (Logothetis & Leopold, 1995). On the other hand, Gorea and Lorenceau (1984) did not find “any systematic dependence on the involuntary eye movements of the observer” during the presentation of supra-threshold counterphase gratings that elicited a strong drift perception. Therefore, whether or not involuntary eye movements are correlated with perception might depend on the particular set of parameters chosen in stimulus generation. However, it is also possible that gratings that are not “optimal” OKN-eliciting stimuli nevertheless induce some statistical fluctuations in perceptual systems that may not always be strong enough to cause supra-threshold activation of the eye-movement system. In this case, at least some sort of statistical relationship should be observable. Recent advances in both eye-tracking technology and automated data-analysis procedures enabled us to detect OKN-like “rudiments” even during visual fixation. 
Microsaccade rate aligned to response changes
The main result with respect to the perceptual consequences of fixational eye movements is the temporal correlation of microsaccade rate and perceptual transitions. Microsaccade rate changes before (reports of) perceptual flips, and it only does so when the detection of perceptual flips is task relevant. Therefore, manual responding alone cannot have caused the rate modulations. Instead, it is likely that they are correlated with the perceptual decision. The timing of microsaccadic inhibition supports this interpretation: in the unambiguous motion condition, it follows changes in motion direction and precedes changes in the response. Here the direction of causality is clear, of course. However, it is unclear whether perceptual flips in the ambiguous condition always precede or sometimes also follow changes in fixational eye movement statistics. In other words, could microsaccades sometimes be causal for perceptual flips? A follow-up experiment was designed to more closely investigate the temporal relationship between perceptual flips and microsaccades. 
Experiment 2
Having established that changes in the frequency of microsaccades are temporally preceding reports of perceptual flips, we were interested in a possible causal role of fixational eye movements on the state of perception: will microsaccades show a similar pattern as LIP activity reported by Williams et al. (2003)? Specifically, we asked if a microsaccade made shortly before the onset of an ambiguous apparent motion stimulus might increase the chance that motion be perceived in a direction systematically related to that of the microsaccade. We tried to answer this question with a second experiment. 
In this experiment, we also investigated whether a possible influence of microsaccades on the perceived direction of motion is modulated by attention. Microsaccadic eye movements have been shown to accompany orienting of spatial attention, at least in situations in which fixation is required. We used this correlation to determine whether we can induce a perceptual bias by presenting attention-capturing peripheral cues. It is known that a flash cue exogenously attracts attention (Theeuwes, 1991; Theeuwes, Kramer, Hahn, Irwin, & Zelinsky, 1999; Yantis & Jonides, 1984), and that microsaccade orientation is affected by attention shifts (Engbert & Kliegl, 2003; Hafed & Clark, 2002). Specifically, microsaccades occurring very shortly after cue presentation tend to be attracted by the exogenous cue, and microsaccades occurring a little later tend to point away from the cued location; the latter effect peaking at about 350 ms after cue onset (Laubrock et al., 2005). Because any sudden-onset cue also inhibits (micro-)saccades, the numeric difference of the cue-opposing effect is greater. To determine the influence of attention, we presented a peripheral sudden-onset cue on a fraction of trials with a fixed SOA of 350 ms before motion onset. The cue was uninformative with respect to the direction of motion and subjects were told to ignore it. The rationale was that if a bias in microsaccade direction can be induced at the time of motion onset, then this bias might affect the perceived motion direction of the ambiguous stimulus. 
Method
Participants
Nineteen female and four male members of the University of Potsdam subject pool (mean age: 23.8, range 19–29 years) were paid seven euros per session for their participation in two experimental sessions lasting for about 45 minutes each. 
Apparatus and materials
The apparatus was the same as in Experiment 1. Also, the stimulus details were similar, with aperture, fixation spot, and spatial grid properties identical. However, vertical grid motion was tested in addition to horizontal motion. Only two motion ambiguity conditions were tested (ambiguous vs. non-ambiguous); in comparison to Experiment 1, the fast-ambiguous and the intermediate ambiguity condition were dropped. One further difference in comparison to Experiment 1 was somewhat larger temporal jitter in the event-related analyses, because due to a programming oversight, synchronization of the eye tracker and computer time stamps was based on the system clock rather than on the vertical retrace. Display changes themselves were still based on the retrace clock. 
The aperture was surrounded by four white circles (1.2 deg diameter), centered on points 10.5 deg above, below, to the left and to the right of fixation, which represented the four possible cue locations, at which uninformative peripheral sudden-onset cues were presented before motion onset on two thirds of the trials. 
At the beginning of each trial, a drift correction was performed using a drift correction target with visual properties identical to the fixation spot. If the drift correction was unsuccessful, the eye tracker was re-calibrated. Scheduled re-calibrations were performed every 32nd trial. If the drift correction was successful, an initial fixation screen was displayed. The fixation screen showed a static grating inside the aperture. 
Following a fixation interval of 1000 ms, on trials with irrelevant cues, one of the outer circles was filled for 50 ms. At 1350 ms into the trial (i.e., 300 ms after cue offset on cued trials), the grid started to move for 1200 ms. After the end of the movement period, a dialog box appeared in the center of the screen, asking for a judgment of movement direction. Responses were given by clicking the appropriate button with a mouse. To avoid a possible influence of manual response preparation on microsaccade-motion congruencies, response buttons were arranged orthogonal to the global direction of grid motion, i.e., “left” above “right” on trials with horizontal motion and “up” next to “down” on trials with vertical motion. 
Design
Participants were tested in two sessions differing in “global” motion direction (horizontally vs. vertically moving grids), with order balanced across participants. On half of the trials within a session, “local” motion was to the left (or up, depending on global motion direction) and on the other half to the right (or down). Motion speed was kept constant at approximately 3 deg/s; therefore, the different levels of ambiguity differed in the degree of displacement of the grid between frames. In the ambiguous condition, the grid was displaced by 0.4 degrees between frames. In the unambiguous condition, the grid was displaced by less than half the dot spacing (0.35 deg). Two cue locations were used per session (left, right or top, bottom). Left and right cues were used in conjunction with horizontal motion, and top/bottom cues in conjunction with vertical motion. On one third of the trials, no cue at all was shown. In the case of unambiguous motion, cue location and movement direction corresponded on half of the remaining two thirds of trials and were different on the other half. The same cue locations were also chosen for ambiguous motion trials, for which the label “correspondence” is not applicable. A session consisted of 240 experimental and 6 warm-up trials. 
Data analysis
Trials containing eye blinks or saccades larger than 1 degree were discarded. Overall, this conservative criterion led to the rejection of 4568 of 11040 trials (mean of 99 discarded trials per session and subject, SD = 48.8). Microsaccade detection and classification followed the same procedures as in Experiment 1
Microsaccade rates were determined by summing across motion-onset aligned trials per condition to give average rates per participant, which were then averaged across participants. Congruencies were determined relative to the response, e.g., on a trial with perceived upward motion, an upward microsaccade was labeled response-congruent and a downward microsaccade response-incongruent. Note that perceived direction was only reported after motion had finished but that a microsaccade could be labeled response-in/congruent even if it occurred before the motion started. 
For trials featuring an irrelevant cue, the same classification scheme was also applied for congruencies between cue location and microsaccade direction, generating microsaccade-cue congruencies. 
Results
Microsaccade rate
As expected, microsaccade rate was influenced by onsets (see Figure 5). Cue and motion onsets both caused inhibition of microsaccade rate. Interestingly, motion onset-related inhibition occurred later on uncued than on cued trials, possibly indicating a preparatory effect of the cue (motion-onset-related inhibition was significantly stronger for cued than uncued trials about 130 ms after motion onset, although no differences in microsaccade rate between cue conditions were obtained at motion onset, see Table 4). During the motion interval, microsaccade rate eventually recovered. The rate peak following motion-onset inhibition was higher for unambiguous than for ambiguous motion. To statistically evaluate these qualitative impressions, we conducted a series of 2-way ANOVAs of microsaccade rate, with factors “cue type” and “ambiguity” on successive, non-overlapping intervals of width 87.5 ms (1/4 of the cue-motion SOA) starting 700 ms before and ending 1400 ms after motion onset. Table 4 shows intervals with significant effects. Because during any interval, only either of the factors was significant, and both the other factor and the interaction had F values smaller than 1, only the statistics for the significant factor in a given interval are listed. 
Figure 5
 
Microsaccade rate in Experiment 2, by cue (present, absent) and motion condition (ambiguous, unambiguous). Rates are locked to motion onset at 0 ms. Both cue presentation and motion onset cause a drop in microsaccade rate. The motion-onset-related drop occurs earlier on cued than on uncued trials, suggesting a preparatory effect of the cue. During the motion interval, the rate of possibly stimulus-triggered microsaccades is higher with unambiguous than with ambiguous motion.
Figure 5
 
Microsaccade rate in Experiment 2, by cue (present, absent) and motion condition (ambiguous, unambiguous). Rates are locked to motion onset at 0 ms. Both cue presentation and motion onset cause a drop in microsaccade rate. The motion-onset-related drop occurs earlier on cued than on uncued trials, suggesting a preparatory effect of the cue. During the motion interval, the rate of possibly stimulus-triggered microsaccades is higher with unambiguous than with ambiguous motion.
Table 4
 
Results of microsaccade rate ANOVAs. Only windows with significant effects are shown (see text for details). Columns “t0” and “t1” list the start and end of the intervals relative to motion onset. The source of the effect is listed in the “factor” column.
Table 4
 
Results of microsaccade rate ANOVAs. Only windows with significant effects are shown (see text for details). Columns “t0” and “t1” list the start and end of the intervals relative to motion onset. The source of the effect is listed in the “factor” column.
t0 t1 Factor F(1,88) p
−262.5 −175.0 Cue 39.37 <.001
−175.0 −87.5 Cue 27.44 <.001
0.0 87.5 Cue 4.57 0.035
87.5 175.0 Cue 10.59 0.002
437.5 525.0 Ambiguity 5.91 0.017
525.0 612.5 Ambiguity 10.05 0.002
612.5 700.0 Ambiguity 8.48 0.005
Congruency of cue location and microsaccade direction
Based on earlier reports, we had expected to be able to bias the distribution of microsaccade directions at motion onset against the cued direction by presenting a peripheral flash cue 350 ms before motion onset. To evaluate the bias, we compared by t-tests the difference in number of cue-congruent and cue-incongruent microsaccades on cued trials in four intervals of 200 ms duration, (1) an interval very shortly after cue presentation (25 to 225 ms after cue onset), where we had previously found a prevalence of cue-congruent microsaccades (Laubrock et al., 2005), (2) the critical interval, centered around motion onset (250 to 450 ms after cue onset), (3,4) two baseline intervals before cue presentation (450 to 250 and 225 to 25 ms before cue onset). 
Results are as expected ( Figure 6). Whereas microsaccades occurring very soon after the cue (25 to 225 ms) were biased toward the cue ( MD = 1.78; t(22) = 2.31; p = 0.045), microsaccades occurring around the time of motion onset tended to be directed away from the cue ( MD = 1.97; t(22) = 2.36; p = 0.016). During none of the corresponding pre-cue intervals were there any biases related to cue location (both MDs < 0.1, both ts < 1, both ps > 0.9). Thus, we replicate our earlier results (Laubrock et al., 2005) and extend them to task-irrelevant cues, for which Galfano, Betta, and Turatto (2004) had previously found only a late tendency away from the cued location but no (significant) early congruent effect. Critically, the cue had the intended effect of biasing microsaccade direction around the time of motion onset. 
Figure 6
 
Congruency of cue location and microsaccade direction in different time intervals (200 ms width, represented by their centers) relative to cue onset in Experiment 2 (see text for details).
Figure 6
 
Congruency of cue location and microsaccade direction in different time intervals (200 ms width, represented by their centers) relative to cue onset in Experiment 2 (see text for details).
Congruency of microsaccades and perceived motion direction
Is microsaccade direction related to motion perception? Is this relation purely stimulus-driven, or does microsaccade-related oculomotor activity also bias motion perception? To answer these questions, we evaluated the average relationship of microsaccade direction for microsaccades occurring in specified intervals to the decision reported at the end of the trial. All times are relative to motion onset, and FDR-corrected statistics are based on non-overlapping intervals of 200 ms width. Because the starting point of these intervals is arbitrary, we re-evaluated the statistics by sliding the starting point across the 200 ms to get a more precise estimate of the location of significant effects. 
Figure 7 depicts the course of microsaccade-decision congruency over the trial for the different cue and motion conditions. A series of t-tests (across subjects) was performed to test whether the difference of decision-congruent and -incongruent microsaccades is different from zero. 
Figure 7
 
Congruency of perceived direction (reported after motion stopped) and microsaccade direction is plotted as a function of time of microsaccade occurrence relative to motion onset. Mean frequency (normalized to per-panel starting value of 1) of decision-congruent (green) and decision-incongruent (red) microsaccade is plotted as a function of time from motion onset, with separate panels for the experimental conditions (columns from left to right: uncued, cued; rows from top to bottom: ambiguous, unambiguous). Microsaccade direction before motion onset (vertical line at x = 0) predicts the perceived direction of ambiguous motion (uncued). If a distracting cue is presented before motion onset (vertical line at x = −350), this influence of microsaccades on perception is destroyed.
Figure 7
 
Congruency of perceived direction (reported after motion stopped) and microsaccade direction is plotted as a function of time of microsaccade occurrence relative to motion onset. Mean frequency (normalized to per-panel starting value of 1) of decision-congruent (green) and decision-incongruent (red) microsaccade is plotted as a function of time from motion onset, with separate panels for the experimental conditions (columns from left to right: uncued, cued; rows from top to bottom: ambiguous, unambiguous). Microsaccade direction before motion onset (vertical line at x = 0) predicts the perceived direction of ambiguous motion (uncued). If a distracting cue is presented before motion onset (vertical line at x = −350), this influence of microsaccades on perception is destroyed.
Results are clear-cut. For unambiguous motion, periods in which microsaccade direction is significantly related to the response last from 236 … 436 until 848 … 1048 ms for uncued trials and from 260 … 460 until motion offset on cued trials. Because the period starts late in the trial, well after motion onset, it is clearly driven by the unambiguous stimulus motion. The direction of the correlation is negative, i.e., microsaccades observed during that interval tend to point in the opposite direction of perceived motion (and hence resemble the fast phase of OKN). Presentation of a cue delays the onset of a significant relationship between microsaccade and response by roughly 40 ms but also prolongs the interval during which it is observed. 
Most importantly, on uncued ambiguous motion trials, microsaccade direction in an interval before motion onset is correlated with the perceived direction of motion, with all tests for intervals between −364 … −164 ms and −308 … −108 ms exceeding the FDR criterion. The direction of the correlation is again negative. Note that this interval coincides with the time at which the cue would have been presented on cued trials. There is also a second, shorter period with a significant negative relationship to the decision, starting at 322 … 522 and ending at 364 … 564 ms after motion onset. 
A Wilcoxon one-sample signed rank test was used as an alternative evaluation of the consistency of this effect across participants. The test statistic was computed by counting the number of subjects for which the difference between the number of response-congruent and response-incongruent microsaccades in the final 800 ms before motion onset was negative. The result again shows that the perceptual bias on motion perception correlated with pre-motion microsaccades is significant across subjects, p = 0.026. The ambiguous uncued condition is the only one with a significant effect. Thus, if microsaccades occur around or before motion onset on uncued trials with ambiguous motion, they influence (or at least indicate) what direction of motion will be perceived. 
For cued ambiguous motion, there appears to be a negative correlation between the direction of microsaccades occurring late in the trial and the perceived direction; however, this is not significant after FDR. More importantly, presentation of the cue destroys the relationship between pre-motion-onset microsaccades and the perceived direction of motion. On cued ambiguous-motion trials, the microsaccade-decision congruency effect fails to reach significance, and numerically, it is even in the opposite direction than on uncued trials. 
Remember that there were significant microsaccade-cue congruency effects. We speculate that source confusion might be an explanation for the different relationship between microsaccades and perception on cued and uncued trials. On cued trials, oculomotor activity does not influence the perceived direction of motion in the same way as on uncued trials because the cognitive system can attribute existing activity as related to an external source. Salient, bottom-up, sudden-onset cues are known to exogenously attract attention and the eye (Laubrock et al., 2005; Theeuwes et al., 1999; Yantis & Jonides, 1984). Therefore, they are likely to induce cue-correlated oculomotor activity and can be attributed to be the source of this activity by higher-level, interpretative systems. If an external source of activation is readily available, it need not be mistakenly attributed to the motion stimulus. 
Percept stabilization
On ambiguous trials, there also exist other perceptual biases. For example, we observe a fairly strong tendency for percept stabilization. Normally, percept stabilization describes the fact that the perceptual switches usually observed during prolonged viewing of an ambiguous stimulus can be slowed by intermittent removal of the stimulus (Leopold, Wilke, Maier, & Logothetis, 2002). In the current context, similarly, the motion pattern is removed during the response phase of the trial. 
We checked whether percept stabilization or priming effects can be found and whether they might be the cause for the oculomotor bias on perception. Descriptively, we observe that what is perceived on trial n tends to be the same as what has been perceived on trial n − 1 ( Table 5). To formally evaluate this dependency, using the package lme4 (Bates, 2007) in the R Environment for Statistical Computing (R Development Core Team, 2008), a linear mixed effects (lme) logistic regression models was specified with response repetition (yes, no) as the criterion, participants as random factor, and motion type (unambiguous vs. ambiguous) and cue type (uncued vs. cued) as fixed effects. Indeed, there is a fairly strong tendency to perceive (or report) the same as has been perceived on the preceding trial (estimate = 0.17, z = 2.61, p = 0.009). Although this percept stabilization (or response repetition) bias is of course stronger on ambiguous than on unambiguous trials (estimate = 0.71, z = 10.32, p < 0.001), it is nevertheless also detectable on the latter (estimate = 0.17, z = 2.82, p = 0.005). On ambiguous trials only, the strong repetition bias (estimate = 0.88, z = 12.24, p < 0.001) is weakened by the cue (estimate = −.16, z = −2.53, p = 0.013), consistent with Kanai and Verstraten (2006), suggesting that distraction of attention interferes with the build-up or maintenance of perceptual memory for stabilization. 
Table 5
 
Percept stabilization. Absolute and relative (with respect to all trials) frequencies of joint occurrence of a response in the current trial and the same vs. the opposite direction as in the previous trial illustrate the dependency of the percept on history, particularly with ambiguous stimulation.
Table 5
 
Percept stabilization. Absolute and relative (with respect to all trials) frequencies of joint occurrence of a response in the current trial and the same vs. the opposite direction as in the previous trial illustrate the dependency of the percept on history, particularly with ambiguous stimulation.
Cue Ambiguity Response same as in last trial?
Same Different
Uncued Unambiguous 1000 (0.09) 840 (0.08)
Ambiguous 1308 (0.12) 532 (0.05)
Cued Unambiguous 1938 (0.18) 1742 (0.16)
Ambiguous 2476 (0.22) 1204 (0.11)
Does the percept stabilization bias interact with the congruency effect between microsaccade direction and direction on ambiguous motion trials? Table 6 shows that this is not the case. On uncued trials, the direction of the incongruency effect is the same for stabilized percepts as for percept changes, and statistical independence cannot be rejected for this sub-table (chi-square = 0, df = 1, p = 0.99). On cued trials, the congruency effect is numerically stronger for percept changes than for stabilized percepts, but again the direction is the same and the sub-table association measure suggests independence (chi-square = 0.52, df = 1, p = 0.47). The significant test of the full table (chi-square = 12.64, df = 4, p = 0.01) indicates however a complex interplay between the two biases. This might be related to the cue, which appears to reduce percept stabilization as well as microsaccade-response congruency. 
Table 6
 
Partitioning of the number of microsaccades in the period before motion onset (starting at −800 ms) on ambiguous motion trials with respect to cue, congruency between microsaccade direction and response, and percept stabilization.
Table 6
 
Partitioning of the number of microsaccades in the period before motion onset (starting at −800 ms) on ambiguous motion trials with respect to cue, congruency between microsaccade direction and response, and percept stabilization.
Microsaccade direction and response Response same as last trial
Same Different
Uncued Incongruent 218 76
Congruent 186 65
Cued Incongruent 323 145
Congruent 343 170
Prolonged pursuit or adaptation to a moving stimulus can lead to an afternystagmus that can influence perception, as it has been shown to represent an extra-retinal influence on the motion aftereffect (e.g., Chaudhuri, 1991; Freeman, Sumnall, & Snowden, 2003). Since we required fixation instead of ocular tracking, and the motion interval was relatively short, the small but systematic deviations from fixation might have induced some sort of microscopic afternystagmus. If so, the previous trial might cause afternystagmus during the stationary phase of the next trial, which in turn might influence the current percept via percept stabilization or priming effects. 
To evaluate this possibility, we analyzed the relationship of the direction of consistent drift during the stationary phase of the trial with the percept reported on the previous trial. We used an lme logistic regression models with participants as random factor and motion type (unambiguous vs. ambiguous) and cue type (uncued vs. cued) as fixed effects to predict congruency of the last or current trial's response (congruent, incongruent) with the baseline pursuit direction. With respect to the afternystagmus hypothesis, we found no evidence for a relationship of any kind between systematic drift during the stationary phase of the trial and percept of the previous trial (all ∣ z∣ < 1, all ps > 0.6). It is important to note that this results holds although percept stabilization was obtained (see above). To conclude, it appears as if the percept stabilization bias does not affect systematic drift or microsaccades in the baseline interval of the trial. Instead, the biases seem to act independently. 
Congruency of systematic drift (OKN slow phases) and perceived motion direction
The “consistent drift” measure is not invalid, however. A further lme analysis reveals that during the motion phase of the trial, there is some pursuit that is systematically related to the percept on the current trial (Intercept: estimate = 1.10, z = 7.75, p < 0.001), and this effect is weaker with ambiguous than unambiguous motion (motion type: estimate −0.69, z = −7.72, p < 0.001). Cue presentation does not affect this relationship ( z = −0.4, p > 0.68). Since pursuit direction is positively correlated with the percept, this can be thought of as the pursuit-counterpart of the microsaccadic effect visible in the right half of each panel of Figure 6. Separate analyses for levels of motion type show that the effect is significant for each ambiguous motion (estimate = 0.40, z = 3.38, p < 0.001) and unambiguous motion (estimate = 1.20, z = 5.35, p < 0.001). 
To summarize qualitatively: like in Experiment 1, a strong tendency for slow fixational eye movement (drift resembling slow pursuit) develops around 400 ms after motion onset. The direction of these slow eye movements is more often biased into the perceived direction of motion than the opposite direction, indicating that it is perception-rather than purely stimulus-related. 
Summary and discussion
The most important result from Experiment 2 is that the direction of eye movements before motion onset is systematically related to the direction of motion that will subsequently be perceived. Thus, it appears as if under ambiguous stimulations, higher-level systems mistakenly attribute the activity in oculomotor systems (as reflected in fixational eye movements) to the stimulus. Part of this drift- and microsaccade-related activity is likely related to the prevention of perceptual fading. 
Alternatively, the changes in the visual input brought about by fixational eye movements are sufficient to disambiguate the ambiguous stimulus. For example, imagine a microsaccade occurring before motion onset in the uncued horizontal ambiguous motion condition. Given the current display arrangement, an average-sized microsaccade starting from the display center would tend to land still within the center grid square but shift the gaze closer to a grid column. When motion starts, the next frame shifts the grid by half the dot distance. Hence, the two most foveal dots would unambiguously travel across the fovea, which might generate a sensation of motion in the corresponding direction. We failed to detect such an effect in our data (by taking into account the post-microsaccadic fixation position with respect to the grid points), however, that failure might be due to limited spatial resolution or suboptimal calibration of the eye tracker. Research using actual stimuli that mimic the effects of the average eye movement response (Rucci, Iovin, Poletti, & Santini, 2007) might provide more decisive results. 
In either case, however, this bottom-up influence of eye movements can be overridden by top-down control as instantiated by the attention- and oculomotor-capturing cue. We use the label “top-down” because we think that despite the exogenous nature of the cue, by the time the decision is reached it is likely to have been evaluated by endogenous processes (e.g., Müller & Rabbitt, 1989). It is known that perceptual rivalry is more readily affected by voluntary attentional control than binocular rivalry (Meng & Tong, 2004), and a voluntary effort to counteract an obvious distractor might have the side effect to reset the bias expressed by earlier microsaccades. The observed temporal modulations in cue-response congruency are consistent with such a speculation. 
Conclusions
We have investigated fixational eye movements during the display of apparent motion stimuli. Specifically, we were interested in whether fixational eye movements such as microsaccades or drift would influence the perceived direction of ambiguous motion. We have shown that eye movements during fixation can both influence and be influenced by the perceptual state. 
A temporal correlation between eye movements and perceptual flips has also been reported by van Dam and van Ee (2006b). These authors reported hints of a causal influence of (micro-)saccades on perception only for binocular rivalry, whereas they found that in both binocular and perceptual rivalry (slant rivalry and Necker cube) a perceptual switch was closely followed by a drop in saccade rate. 
Here we present evidence that such a correlation also exists for perceptual rivalry during ambiguous apparent motion. Changes in microsaccade frequency occur in very close temporal succession to changes in perceptual state. In Experiment 1, these changes clearly preceded reports of the switch. A very similar result has recently been presented by Einhäuser, Stout, Koch, and Carter (2008), who using an ambiguous plaid stimulus not only report an increase of pupil diameter before perceptual switches and a positive correlation of pupil diameter with the dominance duration of the subsequent episode, but also that saccade frequency drops ca. 270 ms before the reported switch and increases afterward (max ca. 450 ms after). 
Although in Experiment 1, the drop in microsaccade rate preceded reports of the switch, it cannot be ruled out that they temporally followed the perceptual event. However, results of Experiment 2 further suggest that at least sometimes what will be perceived later is influenced by microsaccades that occur earlier. This microsaccade-related bias appears to be independent of percept stabilization, and its temporal order is at least compatible with a causal explanation. Nevertheless, we only report a correlation, and microsaccade and perception might well have a common origin. Therefore, “causality” in the following should only be taken to mean temporal precedence (or antecedence). 
Our result that microsaccades can exert a causal influence on the perceptual state goes along well with findings by Martinez-Conde et al. (2006), who reported that the probability and magnitude of microsaccades was decreased before transitions to perceptual fading, and increased before transitions toward visibility. Our results are also in line with findings reported by Kanai et al. (2005), showing that perceptual alternations of several different types of ambiguous stimuli (including apparent motion) could be triggered by a transient stimulus presented nearby. We have shown earlier that such transient stimuli have a systematic effect on microsaccade frequency and direction. Furthermore, eye movements result in a transient signal due to new retinal input. Van Dam and van Ee (2006a) suggested that such a visual transient may contribute to or even be responsible for perceptual alternations. 
Williams et al. (2003) reported a correlation of single-cell responses in the lateral intraparietal area (LIP) of the monkey with the perceived direction of ambiguous apparent motion. However, these authors did not find any systematic effects in fixational eye movements. We observed a correlation of fixational eye movements and the perceived direction of motion in a task closely modeled after Williams et al. (2003), but using a different analysis scheme, concentrating on time-varying and event-related effects. How is this to be interpreted? We have presented evidence that eye movements before motion onset were (statistically) predictive of perceived direction, just like LIP activity reported by Williams et al. (2003). Thus, it appears that either could be causal for the change in perception. However, the direction of causality between eye movements and LIP activity remains an open question. There are several possibilities: First, LIP activity might both influence perceptual decisions and cause a tendency for microsaccades and pursuit to occur. Second, systematic eye movements may be elicited by the stimulus (see also Deubel & Elsner, 1986) and may then cause LIP activity changes correlated with perceptual flips. In other words, if fixational eye movements are correlated with perceived direction, this leaves open the possibility that the neuronal correlations reported by Williams et al. (2003) were at least partly caused by eye movements. Third, changes in both microsaccades and LIP activity might be caused by a common influence, possibly prefrontal activity, which is known to play an important role in determining the perception of ambiguous stimuli (Lumer & Rees, 1999; Sterzer & Kleinschmidt, 2007). In either case, the present results constitute evidence for a relationship between LIP activity and microsaccades. 
Given that fixational eye movements are known to have systematic effects on early motion perception in area MT (Bair & O'Keefe, 1998), our results suggest that small eye movements might be one source of stochastic activity in LIP. Although microsaccades can be suppressed for some time (Winterson & Collewijn, 1976), under normal circumstances they do occur regularly during fixational tasks to prevent visual fading (Engbert & Kliegl, 2004; Engbert & Mergenthaler, 2006; Martinez-Conde et al., 2006). Thus, to the very least, fixational eye movements have a fundamental reason to occur at more or less regular intervals. Furthermore, their timing is stochastic (Engbert, 2006; Mergenthaler & Engbert, 2007), and the unimodal and right-skewed shape of the distribution of inter-microsaccade intervals observed in fixation tasks is similar to the shape of distribution of periods between perceptual alternations. The latter has often been compared to a gamma distribution, although recent studies show that the rate rather than the duration of perceptual alternations follows a gamma distribution (Brascamp, van Ee, Pestman, & van den Berg, 2005). In the present Experiment 1, the distribution of inter-microsaccade intervals can better be fit by a lognormal than by a gamma distribution, suggesting that it is the product of many small independent factors. In the absence of a physically unambiguous stimulus interpretation, “stochastic” fluctuations of LIP activity caused by fixational eye movements might bias the perceptual decision. 
How can microsaccades—events localized in time over a few milliseconds—drive perceptual decisions? The answer is that the movement direction signal carried by microsaccades is temporally more distributed because microsaccades are embedded in and correlated with the drift component of fixational eye movements (Engbert & Kliegl, 2004; see also Engbert, 2006). Recent analyses demonstrated that fixational eye movements are persistent, i.e., they have a tendency to sustain their current movement direction, over a period of about 70 ms (Mergenthaler & Engbert, 2007). Therefore, the accumulation of sensory information for a judgment of movement direction might be based on a time interval much longer than the duration of the microsaccade itself. 
While the determination of the direction of causality between LIP activity and microsaccades requires more stringent experiments than presented here, our findings clearly support the hypothesis that eye movements can induce perceptual alternations and may resolve perceptual ambiguity. 
Acknowledgments
This research was supported by the Deutsche Forschungsgemeinschaft, DFG: KL-955/3. We thank two anonymous reviewers for helpful comments. 
Commercial relationships: none. 
Corresponding author: Jochen Laubrock. 
Email: laubrock@uni-potsdam.de. 
Address: Karl-Liebknecht-Str. 24-25, DE-14415 Potsdam. 
References
Bair, W. O'Keefe, L. P. (1998). The influence of fixational eye movements on the response of neurons in area MT of the macaque. Visual Neuroscience, 15, 779–786. [PubMed] [CrossRef] [PubMed]
Barash, S. Bracewell, R. M. Fogassi, L. Gnadt, J. W. Andersen, R. A. (1991). Saccade-related activity in the lateral intraparietal area: I Temporal properties; comparison with area 7a. Journal of Neurophysiology, 66, 1095–1108. [PubMed] [PubMed]
Bates, D. (2007). lme4: Linear mixed-effects models using S4 classes. R package version 0.99875-9. [Article]
Benjamini, Y. Hochberg, Y. (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society Series B: Methodological, 57, 289–300.
Bisley, J. W. Goldberg, M. E. (2003). Neuronal activity in the lateral intraparietal area and spatial attention. Science, 299, 81–86. [PubMed] [CrossRef] [PubMed]
Brainard, D. H. (1997). The Psychophysics Toolbox. Spatial Vision, 10, 433–436. [PubMed] [CrossRef] [PubMed]
Brascamp, J. W. van Ee, R. Pestman, W. R. van den Berg, A. V. (2005). Distributions of alternation rates in various forms of bistable perception. Journal of Vision, 5, (4):1, 287–298, http://journalofvision.org/5/4/1/, doi:10.1167/5.4.1. [PubMed] [Article] [CrossRef] [PubMed]
Chaudhuri, A. (1991). Eye movements and the motion aftereffect: Alternatives to the induced motion hypothesis. Vision Research, 31, 1639–1645. [PubMed] [CrossRef] [PubMed]
Collins, T. Doré-Mazars, K. (2006). Eye movement signals influence perception: Evidence from the adaptation of reactive and volitional saccades. Vision Research, 46, 3659–3673. [PubMed] [CrossRef] [PubMed]
Cornelissen, F. W. Peters, E. M. Palmer, J. (2002). The Eyelink Toolbox: Eye tracking with MATLAB and Psychophysics Toolbox. Behavior Research Methods, Instruments & Computers, 34, 613–617. [PubMed] [CrossRef]
Deubel, H. Elsner, T. (1986). Threshold perception and saccadic eye movements. Biological Cybernetics, 54, 351–358. [PubMed] [CrossRef] [PubMed]
Deubel, H. Schneider, W. X. (1996). Saccade target selection and object recognition: Evidence for a common attentional mechanism. Vision Research, 36, 1827–1837. [PubMed] [CrossRef] [PubMed]
Einhäuser, W. Stout, J. Koch, C. Carter, O. (2008). Pupil dilation reflects perceptual selection and predicts subsequent stability in perceptual rivalry. Proceedings of the National Academy of Sciences of the United States of America, 105, 1704–1709. [PubMed] [Article] [CrossRef] [PubMed]
Engbert, R. (2006). Microsaccades: A microcosm for research on oculomotor control, attention, and visual perception. Progress in Brain Research, 154, 177–192. [PubMed] [PubMed]
Engbert, R. Kliegl, R. (2003). Microsaccades uncover the orientation of covert attention. Vision Research, 43, 1035–1045. [PubMed] [CrossRef] [PubMed]
Engbert, R. Kliegl, R. (2004). Microsaccades keep the eyes' balance during fixation. Psychological Science, 15, 431–436. [PubMed] [CrossRef] [PubMed]
Engbert, R. Mergenthaler, K. (2006). Microsaccades are triggered by low retinal image slip. Proceedings of the National Academy of Sciences of the United States of America, 103, 7192–7197. [PubMed] [Article] [CrossRef] [PubMed]
Enoksson, P. (1963). Binocular rivalry and monocular dominance studied with optokinetic nystagmus. Acta Ophthalmologica, 41, 544–563. [PubMed] [CrossRef] [PubMed]
Enoksson, P. (1968). Studies in optokinetic binocular rivalry with a new device. Acta Ophthalmologica, 46, 71–74. [PubMed] [CrossRef] [PubMed]
Freeman, T. C. Sumnall, J. H. Snowden, R. J. (2003). The extra-retinal motion aftereffect. Journal of Vision, 3, (11):11, 771–779, http://journalofvision.org/3/11/11/, doi:10.1167/3.11.11. [PubMed] [Article] [CrossRef]
Galfano, G. Betta, E. Turatto, M. (2004). Inhibition of return in microsaccades. Experimental Brain Research, 159, 400–404. [PubMed] [CrossRef] [PubMed]
Gorea, A. Lorenceau, J. (1984). Perceptual bistability with counterphase gratings. Vision Research, 24, 1321–1331. [PubMed] [CrossRef] [PubMed]
Gottlieb, J. (2007). From thought to action: The parietal cortex as a bridge between perception, action, and cognition. Neuron, 53, 9–16. [PubMed] [Article] [CrossRef] [PubMed]
Hafed, Z. M. Clark, J. J. (2002). Microsaccades as an overt measure of covert attention shifts. Vision Research, 42, 2533–2545. [PubMed] [CrossRef] [PubMed]
Hoffman, J. E. Subramaniam, B. (1995). The role of visual attention in saccadic eye movements. Perception & Psychophysics, 57, 787–795. [PubMed] [CrossRef] [PubMed]
Kanai, R. Moradi, F. Shimojo, S. Verstraten, F. A. (2005). Perceptual alternation induced by visual transients. Perception, 34, 803–822. [PubMed] [CrossRef] [PubMed]
Kanai, R. Verstraten, F. A. (2006). Attentional modulation of perceptual stabilization. Proceedings of the Royal Society B: Biological Sciences, 273, 1217–1222. [PubMed] [Article] [CrossRef]
Laubrock, J. Engbert, R. Kliegl, R. (2005). Microsaccade dynamics during covert attention. Vision Research, 45, 721–730. [PubMed] [CrossRef] [PubMed]
Laubrock, J. Engbert, R. Rolfs, M. Kliegl, R. (2007). Microsaccades are an index of covert attention: Commentary on Horowitz, Fine, Fencsik, Yurgenson, and Wolfe (2007. Psychological Science, 18, 364–366. [PubMed] [CrossRef] [PubMed]
Leopold, D. A. Logothetis, N. K. (1999). Multistable phenomena: Changing views in perception. Trends in Cognitive Sciences, 3, 254–264. [PubMed] [CrossRef] [PubMed]
Leopold, D. A. Wilke, M. Maier, A. Logothetis, N. K. (2002). Stable perception of visually ambiguous patterns. Nature Neuroscience, 5, 605–609. [PubMed] [CrossRef] [PubMed]
Logothetis, N. Leopold, D. (1995). On the physiology of bistable percepts.
Lumer, E. D. Rees, G. (1999). Covariation of activity in visual and prefrontal cortex associated with subjective visual perception. Proceedings of the National Academy of Sciences of the United States of America, 96, 1669–1673. [PubMed] [Artcle] [CrossRef] [PubMed]
Madelain, L. Krauzlis, R. J. (2003). Pursuit of the ineffable: Perceptual and motor reversals during the tracking of apparent motion. Journal of Vision, 3, (11):1, 642–653, http://journalofvision.org/3/11/1/, doi:10.1167/3.11.1. [PubMed] [Article] [CrossRef] [PubMed]
Martinez-Conde, S. Macknik, S. L. Hubel, D. H. (2002). The function of bursts of spikes during visual fixation in the awake primate lateral geniculate nucleus and primary visual cortex. Proceedings of the National Academy of Sciences of the United States of America, 99, 13920–13925. [PubMed] [Article] [CrossRef] [PubMed]
Martinez-Conde, S. Macknik, S. L. Hubel, D. H. (2004). The role of fixational eye movements in visual perception. Nature Reviews, Neuroscience, 5, 229–240. [PubMed] [CrossRef]
Martinez-Conde, S. Macknik, S. L. Troncoso, X. G. Dyar, T. A. (2006). Microsaccades counteract visual fading during fixation. Neuron, 49, 297–305. [PubMed] [Article] [CrossRef] [PubMed]
Meng, M. Tong, F. (2004). Can attention selectively bias bistable perception Differences between binocular rivalry and ambiguous figures. Journal of Vision, 4, (7):2, 539–551, http://journalofvision.org/4/7/2/, doi:10.1167/4.7.2. [PubMed] [Article] [CrossRef]
Mergenthaler, K. Engbert, R. (2007). Modeling the control of fixational eye movements with neurophysiological delays. Physical Review Letters, 98, 138104. [CrossRef] [PubMed]
Müller, H. J. Rabbitt, P. M. (1989). Reflexive and voluntary orienting of visual attention: Time course of activation and resistance to interruption. Journal of Experimental Psychology: Human Perception and Performance, 15, 315–330. [PubMed] [CrossRef] [PubMed]
Pelli, D. G. (1997). The VideoToolbox software for visual psychophysics: Transforming numbers into movies. Spatial Vision, 10, 437–442. [PubMed] [CrossRef] [PubMed]
Ramachandran, V. S. Anstis, S. M. (1983). Perceptual organization in moving patterns. Nature, 304, 529–531. [PubMed] [CrossRef] [PubMed]
(2008). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing [[
Rucci, M. Iovin, R. Poletti, M. Santini, F. (2007). Miniature eye movements enhance fine spatial detail. Nature, 447, 851–854. [PubMed] [CrossRef] [PubMed]
Steinman, R. M. Cunitz, R. J. (1968). Fixation of targets near the absolute foveal threshold. Vision Research, 8, 277–286. [PubMed] [CrossRef] [PubMed]
Sterzer, P. Kleinschmidt, A. (2007). A neural basis for inference in perceptual ambiguity. Proceedings of the National Academy of Sciences of the United States of America, 104, 323–328. [PubMed] [Article] [CrossRef] [PubMed]
Theeuwes, J. (1991). Exogenous and endogenous control of attention: The effect of visual onsets and offsets. Perception & Psychophysics, 49, 83–90. [PubMed] [CrossRef] [PubMed]
Theeuwes, J. Kramer, A. F. Hahn, S. Irwin, D. E. Zelinsky, G. J. (1999). Influence of attentional capture on oculomotor control. Journal of Experimental Psychology: Human Perception & Performance, 25, 1595–1608. [PubMed] [CrossRef]
van Dam, L. C. van Ee, R. (2005). The role of (microsaccades and blinks in perceptual bi-stability from slant rivalry. Vision Research, 45, 2417–2435. [PubMed] [CrossRef] [PubMed]
van Dam, L. C. van Ee, R. (2006a). Retinal image shifts, but not eye movements per se, cause alternations in awareness during binocular rivalry. Journal of Vision, 6, (11):3, 1172–1179, http://journalofvision.org/6/11/3/, doi:10.1167/6.11.3. [PubMed] [Article] [CrossRef]
van Dam, L. C. van Ee, R. (2006b). The role of saccades in exerting voluntary control in perceptual and binocular rivalry. Vision Research, 46, 787–799. [PubMed] [CrossRef]
von Helmholtz, H. (1867). Handbuch der Physiologischen Optik. Leipzig: Voss [[
Williams, Z. M. Elfar, J. C. Eskandar, E. N. Toth, L. J. Assad, J. A. (2003). Parietal activity and the perceived direction of ambiguous apparent motion. Nature Neuroscience, 6, 616–623. [PubMed] [CrossRef] [PubMed]
Winterson, B. J. Collewijn, H. (1976). Microsaccades during finely guided visuomotor tasks. Vision Research, 16, 1387–1390. [PubMed] [CrossRef] [PubMed]
Yantis, S. Jonides, J. (1984). Abrupt visual onsets and selective attention: Evidence from visual search. Journal of Experimental Psychology: Human Perception and Performance, 10, 601–621. [PubMed] [CrossRef] [PubMed]
Zhang, M. Barash, S. (2000). Neuronal switching of sensorimotor transformations for antisaccades. Nature, 408, 971–975. [PubMed] [CrossRef] [PubMed]
Zhang, M. Barash, S. (2004). Persistent LIP activity in memory antisaccades: Working memory for a sensorimotor transformation. Journal of Neurophysiology, 91, 1424–1441. [PubMed] [Article] [CrossRef] [PubMed]
Figure 1
 
Characteristics of the moving grid.
Figure 1
 
Characteristics of the moving grid.
Figure 2
 
Distribution of intervals between consecutive microsaccades in Experiment 1. Left: Fit of lognormal, gamma, and inverse gamma distributions to the empirical density (black), smoothed with a Gaussian kernel ( SD 64.8 ms). Right: Empirical vs. fitted 150-tiles in a log–log coordinate system. Both plots show that although there are some systematic deviations, the lognormal evidently fits better than the gamma or inverse gamma distributions.
Figure 2
 
Distribution of intervals between consecutive microsaccades in Experiment 1. Left: Fit of lognormal, gamma, and inverse gamma distributions to the empirical density (black), smoothed with a Gaussian kernel ( SD 64.8 ms). Right: Empirical vs. fitted 150-tiles in a log–log coordinate system. Both plots show that although there are some systematic deviations, the lognormal evidently fits better than the gamma or inverse gamma distributions.
Figure 3
 
Microsaccade rates in Experiment 1; (A) response-locked and (B) stimulus-locked. Microsaccadic inhibition is locked to task-relevant perceptual changes (A). Additionally, every frame-onset causes microsaccadic inhibition with a certain probability (the oscillatory modulation in (B)).
Figure 3
 
Microsaccade rates in Experiment 1; (A) response-locked and (B) stimulus-locked. Microsaccadic inhibition is locked to task-relevant perceptual changes (A). Additionally, every frame-onset causes microsaccadic inhibition with a certain probability (the oscillatory modulation in (B)).
Figure 4
 
Microsaccade rates in the control experiment comparing temporal judgment responses during presentation of a moving vs. static grid, (A) response-locked and (B) stimulus-locked. If changes are not task relevant, no global response-locked microsaccadic inhibition is observed (compare (A) with Figure 3A). However, the moving grid still (as in Figure 3B) elicits frame-locked microsaccadic inhibition, visible in the oscillatory pattern in (B).
Figure 4
 
Microsaccade rates in the control experiment comparing temporal judgment responses during presentation of a moving vs. static grid, (A) response-locked and (B) stimulus-locked. If changes are not task relevant, no global response-locked microsaccadic inhibition is observed (compare (A) with Figure 3A). However, the moving grid still (as in Figure 3B) elicits frame-locked microsaccadic inhibition, visible in the oscillatory pattern in (B).
Figure 5
 
Microsaccade rate in Experiment 2, by cue (present, absent) and motion condition (ambiguous, unambiguous). Rates are locked to motion onset at 0 ms. Both cue presentation and motion onset cause a drop in microsaccade rate. The motion-onset-related drop occurs earlier on cued than on uncued trials, suggesting a preparatory effect of the cue. During the motion interval, the rate of possibly stimulus-triggered microsaccades is higher with unambiguous than with ambiguous motion.
Figure 5
 
Microsaccade rate in Experiment 2, by cue (present, absent) and motion condition (ambiguous, unambiguous). Rates are locked to motion onset at 0 ms. Both cue presentation and motion onset cause a drop in microsaccade rate. The motion-onset-related drop occurs earlier on cued than on uncued trials, suggesting a preparatory effect of the cue. During the motion interval, the rate of possibly stimulus-triggered microsaccades is higher with unambiguous than with ambiguous motion.
Figure 6
 
Congruency of cue location and microsaccade direction in different time intervals (200 ms width, represented by their centers) relative to cue onset in Experiment 2 (see text for details).
Figure 6
 
Congruency of cue location and microsaccade direction in different time intervals (200 ms width, represented by their centers) relative to cue onset in Experiment 2 (see text for details).
Figure 7
 
Congruency of perceived direction (reported after motion stopped) and microsaccade direction is plotted as a function of time of microsaccade occurrence relative to motion onset. Mean frequency (normalized to per-panel starting value of 1) of decision-congruent (green) and decision-incongruent (red) microsaccade is plotted as a function of time from motion onset, with separate panels for the experimental conditions (columns from left to right: uncued, cued; rows from top to bottom: ambiguous, unambiguous). Microsaccade direction before motion onset (vertical line at x = 0) predicts the perceived direction of ambiguous motion (uncued). If a distracting cue is presented before motion onset (vertical line at x = −350), this influence of microsaccades on perception is destroyed.
Figure 7
 
Congruency of perceived direction (reported after motion stopped) and microsaccade direction is plotted as a function of time of microsaccade occurrence relative to motion onset. Mean frequency (normalized to per-panel starting value of 1) of decision-congruent (green) and decision-incongruent (red) microsaccade is plotted as a function of time from motion onset, with separate panels for the experimental conditions (columns from left to right: uncued, cued; rows from top to bottom: ambiguous, unambiguous). Microsaccade direction before motion onset (vertical line at x = 0) predicts the perceived direction of ambiguous motion (uncued). If a distracting cue is presented before motion onset (vertical line at x = −350), this influence of microsaccades on perception is destroyed.
Table 1
 
Statistics (between-subject means and standard errors in milliseconds) for interval between subsequent perceptual flips, interval between subsequent microsaccades, and “reaction time” to changes in stimulus direction in Experiment 1.
Table 1
 
Statistics (between-subject means and standard errors in milliseconds) for interval between subsequent perceptual flips, interval between subsequent microsaccades, and “reaction time” to changes in stimulus direction in Experiment 1.
Condition Perceptual flip waiting time Microsaccades Reaction time
Mean SE Mean SE Mean SE
Unambiguous, 8/20 1468 28.2 1046 105.2 595 33.4
Unambiguous, 9/20 1504 38.6 1126 100.9 672 33.2
Ambiguous 10/20 3077 328.2 1131 102.6 N/A
Ambiguous 10/20, fast 4591 613.1 1113 110.7 N/A
Table 2
 
Frame rate of stimulus changes in the different conditions (Hz) and rate of local modulations of microsaccade rate (Hz).
Table 2
 
Frame rate of stimulus changes in the different conditions (Hz) and rate of local modulations of microsaccade rate (Hz).
Condition Stimulus Microsaccades
Unambiguous, 8/20 8.54 8.54
Unambiguous, 9/20 7.52 7.48
Ambiguous 10/20, normal 6.67 6.62
Ambiguous 10/20, fast 10.00 9.95
Table 3
 
Statistics (between-subject means and standard errors in milliseconds) for intervals between subsequent responses and intervals between subsequent microsaccades in the control experiment.
Table 3
 
Statistics (between-subject means and standard errors in milliseconds) for intervals between subsequent responses and intervals between subsequent microsaccades in the control experiment.
Condition Response alternations Microsaccades
Mean SE Mean SE
Moving grid (ambiguous, 10/20) 945 6.5 926 115.9
No stimulus 985 5.9 849 105.3
Table 4
 
Results of microsaccade rate ANOVAs. Only windows with significant effects are shown (see text for details). Columns “t0” and “t1” list the start and end of the intervals relative to motion onset. The source of the effect is listed in the “factor” column.
Table 4
 
Results of microsaccade rate ANOVAs. Only windows with significant effects are shown (see text for details). Columns “t0” and “t1” list the start and end of the intervals relative to motion onset. The source of the effect is listed in the “factor” column.
t0 t1 Factor F(1,88) p
−262.5 −175.0 Cue 39.37 <.001
−175.0 −87.5 Cue 27.44 <.001
0.0 87.5 Cue 4.57 0.035
87.5 175.0 Cue 10.59 0.002
437.5 525.0 Ambiguity 5.91 0.017
525.0 612.5 Ambiguity 10.05 0.002
612.5 700.0 Ambiguity 8.48 0.005
Table 5
 
Percept stabilization. Absolute and relative (with respect to all trials) frequencies of joint occurrence of a response in the current trial and the same vs. the opposite direction as in the previous trial illustrate the dependency of the percept on history, particularly with ambiguous stimulation.
Table 5
 
Percept stabilization. Absolute and relative (with respect to all trials) frequencies of joint occurrence of a response in the current trial and the same vs. the opposite direction as in the previous trial illustrate the dependency of the percept on history, particularly with ambiguous stimulation.
Cue Ambiguity Response same as in last trial?
Same Different
Uncued Unambiguous 1000 (0.09) 840 (0.08)
Ambiguous 1308 (0.12) 532 (0.05)
Cued Unambiguous 1938 (0.18) 1742 (0.16)
Ambiguous 2476 (0.22) 1204 (0.11)
Table 6
 
Partitioning of the number of microsaccades in the period before motion onset (starting at −800 ms) on ambiguous motion trials with respect to cue, congruency between microsaccade direction and response, and percept stabilization.
Table 6
 
Partitioning of the number of microsaccades in the period before motion onset (starting at −800 ms) on ambiguous motion trials with respect to cue, congruency between microsaccade direction and response, and percept stabilization.
Microsaccade direction and response Response same as last trial
Same Different
Uncued Incongruent 218 76
Congruent 186 65
Cued Incongruent 323 145
Congruent 343 170
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×